KRLS: Improving End-to-End Response Generation in Task Oriented Dialog with Reinforced Keywords Learning
Abstract
In task-oriented dialogs (TOD), reinforcement learning (RL) algorithms train a model to directly optimize response for task-related metrics. However, RL needs to perform exploration, which can be time-consuming due to the slow auto-regressive sequence generation process. We investigate an approach to create a more efficient RL-based algorithm to improve TOD performance in an offline setting. First, we use a faster generation procedure that samples from independent next-word distributions after training the language model (LM) with supervised learning. We then introduce a fine-grained reward function to help the model focus on learning key information in a dialog, by measuring the importance and semantic closeness of each generated token. Experiments on the MultiWoZ dataset show our new training algorithm, Keywords Reinforcement Learning with Next-word Sampling (KRLS), achieves state-of-the-art performance on the end-to-end response generation task, with a 15% training time reduction compared to a standard RL algorithm using auto-regressive generation.
Cite
@article{arxiv.2211.16773,
title = {KRLS: Improving End-to-End Response Generation in Task Oriented Dialog with Reinforced Keywords Learning},
author = {Xiao Yu and Qingyang Wu and Kun Qian and Zhou Yu},
journal= {arXiv preprint arXiv:2211.16773},
year = {2023}
}
Comments
Accepted at EMNLP 2023